personnelSelectionUtility 1.0.2

Second CRAN review resubmission

• Added explicit \value{} documentation for the print.psu_utility() S3 method (and its subclass aliases including print.psu_sturman() and print.psu_comparison()), following the CRAN reviewer’s request. The previous resubmission missed the value tag in the regenerated Rd file due to a stale roxygen block; this has been corrected.

personnelSelectionUtility 1.0.1

CRAN review resubmission

• Added DOI references in the Description field of DESCRIPTION as requested by the CRAN reviewer.
• Added explicit \value{} documentation to print.psu_utility() and its subclass aliases.

personnelSelectionUtility 1.0.0

First stable release.

Models implemented

• Classification utility: Taylor-Russell univariate (tr_classic, tr_solve, tr_binomial_success_probability) and Thomas-Owen-Gunst multivariate extension for conjunctive multiple-hurdle selection (tr_multivariate, tr_multivariate_equal_cutoff, group_tr_multivariate).
• Continuous utility: Naylor-Shine (naylor_shine), Brogden-Cronbach-Gleser (bcg_utility), Schmidt-Hunter-Pearlman intervention utility (shp_utility), Boudreau-style discounted utility (boudreau_utility), probation-period adjustments (probation_adjustment, probation_utility), and Sturman’s (2001) integrated comprehensive model (sturman_comprehensive).
• SDy estimation: percentile method (sdy_percentile), proportional rules (sdy_proportional), Raju-Burke-Normand coefficient of variation (sdy_rbn), CREPID-style activity decomposition (sdy_crepid), cost-accounting (sdy_cost_accounting), superior-equivalents (sdy_superior_equivalents), observed (sdy_observed).
• Incremental validity: restricted canonical validity for a fixed criterion composite (restricted_canonical_validity), incremental validity (incremental_validity), Johnson relative weights (relative_weights), Budescu dominance analysis (dominance_analysis), composite effect size (composite_d).
• Composite formation (Lord & Novick, 1968): composite reliability (fuse_reliability), composite validity (fuse_validity), composite correlation matrix (fuse_composite_cor), Spearman disattenuation (disattenuate_correlation).
• Range-restriction corrections: direct (Thorndike Case II, correct_r_direct_range_restriction) and multivariate Lawley (correct_r_lawley) for incidental restriction.
• Offer-rejection models (Hogarth-Einhorn, 1976; Murphy, 1986): offer_rejection_adjustment with uniform, correlated, and selective modes.
• Selection systems: compensatory top-down (compensatory_selection), conjunctive multiple-hurdle simulation (multiple_hurdle_selection, multiple_hurdle_selection_staged), and Ock-Oswald-style comparisons (compare_selection_systems, compare_selection_systems_staged).
• Decision support: multi-attribute utility (multiattribute_utility), Pareto frontiers (pareto_frontier, utility_fairness_frontier), adverse-impact ratio (adverse_impact_ratio), Monte Carlo uncertainty propagation (utility_monte_carlo), sensitivity grid (sensitivity_grid), break-even validity (break_even_validity), risk-adjusted utility (risk_adjusted_utility).
• Conversions: AUC <-> rank-biserial, AUC -> Cohen’s d under the equal-variance binormal model, d <-> point-biserial correlation, inflation-adjusted discount rate.
• Diagnostics: regression-based linearity and normality checks (utility_regression_diagnostics), forecasting efficiency (forecasting_efficiency), coefficient of determination (coefficient_of_determination).
• Reference helpers: model taxonomy and argument glossary as data frames (model_taxonomy, argument_glossary).

Tests and validation

• The Taylor-Russell univariate integral and the Thomas-Owen-Gunst multivariate integral are cross-validated against scipy.stats.multivariate_normal on a 6 x 6 grid (TR 1939) and a 3 x 3 grid (TOG 1977). Discrepancies are below 1e-3 throughout.
• correct_r_lawley validated to reproduce Thorndike Case II in the k = 2 case to machine precision.
• dominance_analysis validated such that general-dominance values sum to the full-model R^2 to machine precision.
• fuse_reliability validated against the Spearman-Brown formula for parallel items.
• Internal coherence: tr_multivariate(k = 1) matches tr_classic to machine precision; boudreau_utility collapses to bcg_utility under degenerate parameters; cor_to_d and d_to_cor invert each other.

Vignettes

• getting-started: tour through the five main families of the package.
• reproductions: reproduces the qualitative pattern of Sturman’s (2001) comprehensive cascade and Ock and Oswald’s (2018) compensatory vs. multiple-hurdle comparison, plus a worked example with sturman_comprehensive().

Continuous integration

• R-CMD-check.yaml: macOS / Windows / Ubuntu (devel, release, oldrel-1).
• test-coverage.yaml: codecov via covr.
• pkgdown.yaml: site deployment to gh-pages.